Data validity assessment with computer vision

Surveying in optimal conditions produces the most repeatable results. However, optimal conditions are not always possible for practical reasons. To address this, RoadAI uses a dedicated computer vision model to analyze recorded videos for context, recording conditions, and road surface state. These are the factors that impact overall data validity. The purpose of this analysis is to determine whether the collected video material is suitable for extracting specific data.

Data validity assessment affects line data, such as road surface condition and line marking data, when you create reports. In the Data validity settings step of the line data report creation, you can choose values for each factor that contributes to the final definition of valid data for the report. Using the default settings produces the best results. However, you may change these settings if necessary. Changing the default values is not recommended, as this may result in a report that contains data that is not valid or repeatable. You may, however, adjust the defaults for specific use cases, such as when you create a concrete surface condition report.

When you generate a report, the data is always based on the most recent videos that match the selected validity criteria within the applied filters. If more recent videos do not match the validity criteria, they are not used in the report. This approach, which excludes unsuitable video material, is applied at a resolution of 5 meters.